Upgrade NetCDF output writer to write arbitrary outputs for LESbrary use

[ali-ramadhan]

This PR allows NetCDFOutputWriter to write arbitrary outputs, similar to how we pass functions that return data to be saved to disk with the JLD2 output writer, but now with metadata! The dimensions kwarg allows you to pass a dictionary of dimensions to apply to any outputs.

I also cleaned up the NetCDFOutputWriter a bit and changed the name of the time dimension from "Time" to "time".

@suyashbire1 Let me know if I did anything wrong.

I added a pretty comprehensive test that shows how you might write a mix of scalars, profiles, and slices to a NetCDF file with grid/unit/name metadata and global attributes.

Here's what the API looks like right now:

# Define scalar, vector, 2D slice, and 3D field outputs
f(model) = model.clock.time^2
g(model) = @. model.clock.time * exp(model.grid.zC)
h(model) = @. model.clock.time * sin(model.grid.xC) * cos(model.grid.yC')

outputs = Dict("scalar" => f, "profile" => g, "slice" => h)
dims = Dict("scalar" => (), "profile" => ("zC",), "slice" => ("xC", "yC"))

output_attributes = Dict(
    "scalar"  => Dict("longname" => "Some scalar", "units" => "bananas"),
    "profile" => Dict("longname" => "Some vertical profile", "units" => "watermelons"),
    "slice"   => Dict("longname" => "Some slice", "units" => "mushrooms")
)

global_attributes = Dict("location" => "Bay of Fundy", "onions" => 7)

simulation.output_writers[:fruits] =
    NetCDFOutputWriter(
        model, outputs; frequency=1, filename="test_function_outputs.nc", dimensions=dims,
        global_attributes=global_attributes, output_attributes=output_attributes)

run!(simulation)

[glwagner]

Beautiful. Is this documented? Is there an example? The two_dimensional_turbulence.jl example could be appended to output vorticity, perhaps... or one of the three-dimensional examples might be modified.

[ali-ramadhan]

The NetCDF tests fail on GPU so gotta look into that.

Is this documented? Is there an example?

No and no, but yeah we should update model setup documentation and show an example of how to use it. Vorticity example might be good. Might also be nice to have the example save a horizontal average to NetCDF so we can also show how to save outputs with different dimensions.

[codecov]

Codecov Report

Merging #643 into master will decrease coverage by 2.9%.
The diff coverage is 97.22%.

@@            Coverage Diff            @@
##           master    #643      +/-   ##
=========================================
- Coverage   78.11%   75.2%   -2.91%     
=========================================
  Files         119     119              
  Lines        2326    2267      -59     
=========================================
- Hits         1817    1705     -112     
- Misses        509     562      +53

Impacted Files	Coverage Δ
src/OutputWriters/OutputWriters.jl	100% <ø> (ø)	⬆️
src/OutputWriters/checkpointer.jl	90.9% <ø> (-0.62%)	⬇️
src/OutputWriters/netcdf_output_writer.jl	87.5% <100%> (-12.5%)	⬇️
src/OutputWriters/output_writer_utils.jl	61.53% <50%> (-0.63%)	⬇️
src/Coriolis/no_rotation.jl	0% <0%> (-100%)	⬇️
src/Forcing/Forcing.jl	50% <0%> (-50%)	⬇️
src/SurfaceWaves.jl	0% <0%> (-47.06%)	⬇️
src/AbstractOperations/interpolation_utils.jl	63.33% <0%> (-33.34%)	⬇️
src/AbstractOperations/AbstractOperations.jl	33.33% <0%> (-33.34%)	⬇️
src/Forcing/simple_forcing.jl	50% <0%> (-33.34%)	⬇️
... and 38 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9f72d84...7809b1e. Read the comment docs.

[glwagner]

No and no, but yeah we should update model setup documentation and show an example of how to use it. Vorticity example might be good. Might also be nice to have the example save a horizontal average to NetCDF so we can also show how to save outputs with different dimensions.

Maybe just get this merged and open an issue?

[glwagner]

typo

[glwagner]

Typo here. There is also a spurious file called / above.

[glwagner]

@ali-ramadhan a spurious file should be deleted before this is merged.

[ali-ramadhan]

Ah thanks for catching that, no idea how it happened.

[ali-ramadhan]

Maybe just get this merged and open an issue?

Yeah sure, there are a few other things that need updating too.